Identification of emotions in texts of a social network

Abstract

Social networks are often used to express opinions on different aspects
of society, products, services, politics, celebrities, etc. Companies,
organizations and governments have shown interest in knowing what
users think about their activities or products. In addition to determining whether an opinion is positive or negative, it
is interesting to determine what the feeling or
emotion expressed in the opinion is. Identifying the
emotion that a user expresses in a textual message
can be understood as classifying or categorizing the
message according to its characteristics.
In this work, a method was developed to classify
short texts or opinions of the social network
Twitter, according to the emotion they express.
First, it was necessary to structure the texts by
discarding irrelevant parts, but trying to keep as
much information as possible. Then, automatic
learning techniques were used to generate a
corpus of tagged opinions. Finally, a method of
classification by weighting was applied with lexical
dictionaries associated with three emotional
values: valence, activation and dominance.